Computing the Stereo Matching Cost with a Convolutional Neural Network

CVPR 2015

doi:10.1109/CVPR.2015.7298767 /cvpr/2015/zbontar2015cvpr-computing/

Abstract

We present a method for extracting depth information from a rectified image pair. We train a convolutional neural network to predict how well two image patches match and use it to compute the stereo matching cost. The cost is refined by cross-based cost aggregation and semiglobal matching, followed by a left-right consistency check to eliminate errors in the occluded regions. Our stereo method achieves an error rate of 2.61 % on the KITTI stereo dataset and is currently (August 2014) the top performing method on this dataset.

PDF CVPR Semantic Scholar

Cite

Text

Zbontar and LeCun. "Computing the Stereo Matching Cost with a Convolutional Neural Network." Conference on Computer Vision and Pattern Recognition, 2015. doi:10.1109/CVPR.2015.7298767

Markdown

[Zbontar and LeCun. "Computing the Stereo Matching Cost with a Convolutional Neural Network." Conference on Computer Vision and Pattern Recognition, 2015.](https://mlanthology.org/cvpr/2015/zbontar2015cvpr-computing/) doi:10.1109/CVPR.2015.7298767

BibTeX

@inproceedings{zbontar2015cvpr-computing,
  title     = {{Computing the Stereo Matching Cost with a Convolutional Neural Network}},
  author    = {Zbontar, Jure and LeCun, Yann},
  booktitle = {Conference on Computer Vision and Pattern Recognition},
  year      = {2015},
  doi       = {10.1109/CVPR.2015.7298767},
  url       = {https://mlanthology.org/cvpr/2015/zbontar2015cvpr-computing/}
}